关键词: 阴影消除植被指数（SEVI）, 本影, 落影, 地形校正模型, 植被监测, Landsat 5 TM, NDVI

Abstract:

Topographic correction is a crucial step in the pre-processing of remote sensing imagery of rugged terrain areas. Recently, a Shadow-Eliminated Vegetation Index (SEVI) was proposed to eliminate the influence of the self and cast terrain shadows. To further evaluate the SEVI performance for reducing the terrain shadow effect on regional scales, here we compared the SEVI with classic topographic correction models, including the C model, Sun-Canopy-Sensor (SCS)+C model, and Second Simulation of the Satellite Signal in the Solar Spectrum (6S)+C model via the case study of Fuzhou city, China. Landsat 5 TM satellite data and associated 30-meter Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model Version 2 (ASTER GDEM V2) were used for the comparison. The satellite imagery were first corrected using the C, SCS+C, and 6S+C models, followed by the calculation of the Normalized Difference Vegetation Index (NDVI) and the Ratio Vegetation Index (RVI). Then, the calculated vegetation indices were evaluated in different ways, including visual comparison, statistical analysis, and linear correlation analysis of the cosine of solar incidence angle ( \cos i ) versus vegetation indices. The C and SCS+C models showed accurate correction results over the self shadow areas but less accurate results over the cast shadow areas. Using adjacent sunny slopes as a reference, the relative errors of the NDVI and RVI over the self shadow areas were reduced from 71.64% and 52.57% to 4.80% and 6.43% (C model) and 0.50% and 9.94% (SCS+C model), respectively; the relative errors over cast shadow areas were reduced from 62.01% and 47.57% to 31.05% and 24.40% (C model) and 33.42% and 16.01% (SCS+C model), respectively. The 6S+C model showed better correction results over the cast shadow areas than the C model and the SCS+C model did. The relative errors of the NDVI were 8.63% and 14.27% over self shadow and cast shadow areas, respectively, if the 6S+C model was used. The SEVI seemed the most accurate among these models for corrections of self and cast shadows. The relative errors of the SEVI were 9.86% and 10.53% over the self and cast shadows, respectively. Finally, the SEVI was used to study the vegetation cover change in Fuzhou city from 1999 to 2014. Results show: (1) the vegetation cover in Fuzhou city increased from 1999 to 2014 in general, particularly over the areas with elevation ranging from 250 to 1250 meters; (2) The highest SEVI mean was located on the slope of about 40 degrees.

Key words: Shadow-Eliminated Vegetation Index, self shadow, cast shadow, topographic correction model, vegetation monitor, Landsat 5 TM, NDVI